• Laser & Optoelectronics Progress
  • Vol. 61, Issue 8, 0837012 (2024)
Jiaguang Huang1, Zhenming Yu2,*, Guojin Peng2, Hui Gan2, and Lü Meini2
Author Affiliations
  • 1College of Information and Communication, Guilin University of Electronic Technology, Guilin 541004, Guangxi , China
  • 2Guangxi Key Laboratory of Machine Vision and Intelligent Control, Wuzhou University, Wuzhou 543002, Guangxi , China
  • show less
    DOI: 10.3788/LOP230799 Cite this Article Set citation alerts
    Jiaguang Huang, Zhenming Yu, Guojin Peng, Hui Gan, Lü Meini. Microscopic Image Stitching Algorithm Based on Stage Motion Information[J]. Laser & Optoelectronics Progress, 2024, 61(8): 0837012 Copy Citation Text show less
    Motion platform
    Fig. 1. Motion platform
    Stitching path
    Fig. 2. Stitching path
    Predictive stitching model
    Fig. 3. Predictive stitching model
    Predict matching feature points
    Fig. 4. Predict matching feature points
    Feature point pair slope coarse screening
    Fig. 5. Feature point pair slope coarse screening
    Stitching mode
    Fig. 6. Stitching mode
    Registering image with stitching direction of 1
    Fig. 7. Registering image with stitching direction of 1
    Image fusion. (a) Fusion of two images; (b) fusion of multiple images
    Fig. 8. Image fusion. (a) Fusion of two images; (b) fusion of multiple images
    Multi-image stitching
    Fig. 9. Multi-image stitching
    Algorithm flow with stitching direction of 2
    Fig. 10. Algorithm flow with stitching direction of 2
    Similarity of image structure. (a) Spinal cord; (b) dicotyledons; (c) natural scene
    Fig. 11. Similarity of image structure. (a) Spinal cord; (b) dicotyledons; (c) natural scene
    Predicted and actual coordinates of different overlapping areas. (a) 20% image overlap area; (b) 30% image overlap area
    Fig. 12. Predicted and actual coordinates of different overlapping areas. (a) 20% image overlap area; (b) 30% image overlap area
    Experimental results of performance comparison of different matching algorithms. (a) BF first match; (b) Flann first match; (c) proposed algorithm first match; (d) BF good match; (e) Flann good match; (f) proposed algorithm good match
    Fig. 13. Experimental results of performance comparison of different matching algorithms. (a) BF first match; (b) Flann first match; (c) proposed algorithm first match; (d) BF good match; (e) Flann good match; (f) proposed algorithm good match
    Image fusion. (a) Before image fusion; (b) after image fusion
    Fig. 14. Image fusion. (a) Before image fusion; (b) after image fusion
    3×3 images stitching. (a) Stitched original images; (b) AutoStitch stitching; (c) proposed method stitching
    Fig. 15. 3×3 images stitching. (a) Stitched original images; (b) AutoStitch stitching; (c) proposed method stitching
    4×5 images stitching. (a) AutoStitch stitching; (b) proposed method stitching
    Fig. 16. 4×5 images stitching. (a) AutoStitch stitching; (b) proposed method stitching
    Microscopic image stitching system. (a) Motorized stage; (b) (c) part of the process of stitching images in real time
    Fig. 17. Microscopic image stitching system. (a) Motorized stage; (b) (c) part of the process of stitching images in real time
    Type of imageThresholdNumber of feature pointsNumber of similar feature pointsSimilarity /%
    Spinal cord0.2211944821.15
    Dicotyledons0.2318950015.68
    Natural scene0.2803121.49
    Table 1. Structural similarity of different types of images
    Image overlapping area /%Predicted distanceMatch point(xyPredicted coordinate(xyActual matching point(xyX error /%Y error /%Euclidean distance
    20320,01553,746273,746276,7500.190.330.0283
    320,01490,189210,189213,1920.190.250.0667
    320,01334,115854,115857,11600.190.170.0506
    30480,01463,208343,208339,2120.250.330.03584
    480,01416,794296,794292,7960.250.160.04625
    480,01485,303365,303361,3060.250.250.02003
    Table 2. Predicted and actual coordinates of matching points in different overlapping areas
    Images of experimentAlgorithmLeft feature pointsRight feature pointsMatch feature pointsExcellent match pointsMatch rate /%Time consumption /s
    PollenBF8867882528.400.001
    Flann8867882730.680.001
    Ours8867883438.630.001
    Spinal cordBF2319197123192109.050.226
    Flann2319197123192219.520.035
    Ours23191971231982535.570.028
    DicotyledonBF45774379457786918.990.512
    Flann45774379457787519.120.075
    Ours457743794577126327.590.061
    Table 3. Performance comparison of different matching algorithms
    MethodNumber of stitched imagesMicroscope magnificationImage overlapping area /%Time consumption /s
    AutoStitch3×32003012.33
    4×53003030.39
    Ours3×3200302.84
    4×5300309.17
    Table 4. Multi-image stitching
    Jiaguang Huang, Zhenming Yu, Guojin Peng, Hui Gan, Lü Meini. Microscopic Image Stitching Algorithm Based on Stage Motion Information[J]. Laser & Optoelectronics Progress, 2024, 61(8): 0837012
    Download Citation